CFD Lab Department of Engineering The University of Liverpool

Transcription

1

2 Development of a CFD Method for Aerodynamic Analysis of Large Diameter Horizontal Axis wind turbines S. Gomez-Iradi, G.N. Barakos and X. Munduate 2007 joint meeting of IEA Annex 11 and Annex 20 Risø National Laboratory Roskilde, Denmark, June, 2007

3

4

5 CFD Solver - Overview

6 Summary of Features - 1 PDE solver Multi-block capability Parallerised using the SPMD paradigm Flow Physics: Inviscid, RANS, URANS, DES, LES Aeroelastic analysis based on modal representation of structures Moving and deforming grids Modular code with a uniform data structure based on the linked-list concept

7 Summary of Features - 2 Control volume method Parallel - Shared and Distributed memory Multi-block (complex geometry) structured grids Unsteady RANS - Variety of turbulence models inc. LES/DES Implicit time marching Osher's and Roe's schemes for convective fluxes MUSCL scheme for formally 3 rd order accuracy Central differences for viscous fluxes Krylov subspace linear solver with pre-conditioning Moving grids, sliding planes Aeroelastic analysis based on modal representation of structures Hover formulation, rotor trimming, blade actuation Modular code with a uniform data structure based on the linked-list concept Documentation Validation database Range of utilities for processing data, structural models etc. Used by academics and engineers

8 Requirements Minimal requirements ANSI C compiler MPI libraries for parallel computing Grid generation (ICEM CFD at present) Post-processing (Tecplot, Fieldview, Vigie, Paraview) Any Unix flavour will do Optimum (low cost) performance on Beowulf clusters Ported to national computing centres, CSAR, HPCx, IAG

9 Multi-block topologies for rotor cases - ICEMCFD 2-4.5M grid points per blade, blade actuation requires special topologies

10 Add-ons for TECPLOT

11 Probe Analyser - MATLAB

12 Validation Database Distributed with the solver Contains public and restricted cases (sponsorspecific) Public cases are accessible via www Each case is selected to demonstrate and check a particular feature Experimental data are available for all cases Sample results are included

13 Partial List of Validation Cases RAE2822 Case 9 Attached Flow ONERA A Aerofoil Separation under apg Williams aerofoil Multi-element sections NACA0012 & AGARD CT2 Pitching aerofoils, inviscid flow Bachalo-Johnson Bump SBL interaction Delery s Bump C SBL interaction internal flow 2D cavity flow Unsteady turbulent flow Convected vortex Vorticity confinement AGARD wing Inviscid aeroelastic ONERA M6 wing Viscous turbulent flow over a wing NACA wing - UNSI 3D dynamic stall AGARD S-duct Turbulent separated internal flow 65 and 70-degree delta wing Vortical turbulent flow ONERA Rotor 7A/7AD Helicopter rotor flow Glasgow BERP tip Tip flow NACA0015 oscillating wing DS - tip flow LABM low-re DS cases Transitional flow - dynamic stall

14 RAE 2822 Aerofoil

15 Documentation Organised as a set of Technical Notes Important part of the CFD solver At present there are 14 technical notes covering various features of the code Information is included in the source code in the form of comments Under continuous development!

16 3D DS validation - LABM (i) Pitching motion, k = AoA= sin(wt) x/c = 0.4 z/c =0.5 AoA = 18 deg upstroke AoA = 24 deg

17 Blade-Vortex Interaction x/c=0.02

18 CFD Validation - UH-60A Model Main Rotor Tests - Lorber 0.965R 0.775R 0.865R 0.920R 0.945R

19 Sponsors of our work

20 HAWT - Multi-Block Topologies

21 Annex XX Data NASA Ames wind tunnel 24.4 m x 36.6 m test section Two bladed upwind wind turbine, with S809 aerofoil after the 25% of the span Test instrumentation (Input) 22 Pressure taps each at 5 spanwise possitions Accelerometers in both blades (edge & flap) and in the nacelle (yaw, fore-aft & pitch) Strain gauges in both blades (root. edge & flap) Wind tunnel's dynamic, static and total pressures, density, temperature, velocity,... Data files (Output) Raw data & Averaged data (Azimuth & Cycle) Azimuthal and raw average data were used

22 Rigid Blocks around HAWT Blades CFD Lab Department of Engineering The University of Liverpool

23 176 blocks 2.4 million cells Blade boundary 10-5 Case L (Parked and Pitching cases) 30 chords 30 chords 30 c 30 c INFLOW OUTFLOW 1 Radius = 6.8 chords

24 524 blocks 4.55 million cells Blade boundary chords Clean Configuration 30 chords 15 chords Spinner OUTFLOW INFLOW 60 chords ` 1 Radius = 6.8 chords

25 Investigation of the Blade Shape Real Area (aprox.) = CENER Area = LIVERPOOL Area = `

26 First Results

27 Unsteady Navier-Stokes Equations: Continuity equation CFD Lab Department of Engineering The University of Liverpool Steady and Unsteady States Momentum Conservation Energy equation

28 Steady Navier-Stokes Equations: CFD Lab Department of Engineering The University of Liverpool Steady and Unsteady States Continuity equation Non-Inertial frame of reference Momentum Conservation Energy equation No temporal variation, so their derivate respect the time is equal to zero

29 Turbulence Modelling κ-ω of Wilcox D.C Wilcox, Simulation of Transition with a Two-Equation Turbulence Model, AIAA Journal, Vol. 32, No. 2, February 1994

30 L20000 Case

31 L20000 Case Computation Grid Size: 2,416,584 Load Balance: 8 computers 0.2% Convegency: All in 2000 steps to 10-6 Mach number: 0.15 (1) & 0.20 (2) M=0.20 T=3002 Turbulence model: 3000 (2) & 3002 (1) M=0.15 T=3000 M=0.20 T=3000

32 L20000 Case

33 L20000 Case

34 L2000ST0SD Case

35 Computation L2000ST0SD Case Grid Size: 2,416,584 Load Balance: 24 computers 3.7% Convegency: Initially in 2000 steps to 10-6 (Euclidean Convergence) 2-3 for the azimuthal variation (Unsteady Convergence) Mach number: 0.15 Turbulence model: 3000 (1) & 3002 (1)

36 L2000ST0SD Case 46.6 % 70 o pitch 50 o pitch ( o local pitch) ( o local pitch)

37 L2000ST0SD Case 63.3 % 70 o pitch 50 o pitch ( o local pitch) ( o local pitch)

38 L2000ST0SD Case 80 % 70 o pitch 50 o pitch ( o local pitch) ( o local pitch)

39 S Case Steady Computation

40 Computation Grid Size: 4,552,304 S Case Load Balance: 33 computers 0.7 % Mach number: 0.1 (1) Turbulence model: 3000 (1)

41 S Case Steady Computation C P distribution 30 % 46.6 %

42 C P distribution CFD Lab Department of Engineering The University of Liverpool S Case Steady Computation 63.3 % 80 %

43 S Case Integrated Loads

44 S Case Tip vortex Inner vortex ReT λ 2

45 S Case - Unsteady Computations High wind speed and stalled behaviour Reynolds number based in the maximum chord of the blade and the free stream velocity.

46 Grid Size = 6,550,400 (524 blocks) Load Balance = 55 processors (97.7%) Turbulence model: κ-ω of Wilcox S Case (Unsteady State)

47 S Case - Steady Computation Lower Surface (Pressure side) 0 o azimuth angle Steady Unsteady

48 S Case - Steady Computation Upper Surface (Suction side) 0 o azimuth angle Steady Unsteady

49 20m/s

50 20m/s

51 20m/s

52 S Case (Unsteady Flow Solution) λ 2 1 Radius

53 Future Steps

54 Sliding Meshes

55 Free Stream and Wind Tunnel Configurations 60 chords 33.1 chords = 24.4m Free stream configuration 60 chords away from the wind turbine hub 49.7 chords = 36.6m Wind tunnel configuration Dimensions normalised with the maximum chord in the blade 1 Radius = 6.8 chords

56 Rotor, Nacelle and Tower Configurations Blades & Spinner OUTFLOW INFLOW Nacelle & tower 33.1 chords 60 chords 30 chords 49.7 chords

57 Sliding Grids F F F R R R F F F F = Fix R = Rotating

58 Sliding Grids F F F F F F F R R F F F = Fix R = Rotating

59 Sliding Plane Method Formation of regular intermediate planes to avoid general cloud-tocloud interpolations Interpolate from fixed mesh to corresponding intermediate planes Interpolate from rotating mesh to corresponding intermediate planes Set halo-cells on both side of sliding plane using data on intermediate planes Intermediate plane data stored on each CPU Identification on small patches of regular planes

60 Questions?

61